Leahnaya/pokemon-showdown
1
0
Fork 0
mirror of https://github.com/smogon/pokemon-showdown.git synced 2026-05-18 19:28:35 -05:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity
b0901e6fc9
pokemon-showdown/dev-tools/HARNESS.md

2.3 KiB
Raw Blame History

Harness

harness.js allows for running multiple random simulations of Pokemon battles for testing or benchmarking purposes. Without any arguments, harness.js will run 100 random battles and report any errors that occurred. The number of battles run can be configured through by passing the number as the sole argument to harness.js, or through the --num flag (see below).

Flags

Using any flag will trigger minimist to be installed if it has not been already.

General

  • --num: play a specific number of games for a format instead of the default 100.
  • --seed: PRNG seed to use (eg. '1234,5678,9012,3456').
  • --output: makes the harness display the output logs of each battle it runs.
  • --input: dump the battle input logs of each battle it runs.
  • --error: dump the battle input logs of each battle which errors.

Format

By default, the harness will select the format of the next game it runs randomly based on its initial --seed. Alternatively, it can run games all in the same format, cycle through the formats or run all formats.

  • --format: play the specified format for each of the games it runs. Note that the harness only supports formats where the team can be randomly generated.
  • --cycle: cycles through the possible formats, playing one battle in each --num battles in total.
  • --all: plays every supported format --num times before moving on to the next format.

Concurrency

The harness runs games sequentially by default, but can be configured to run games asynchronously.

  • --async: runs each game concurrently instead of waiting for the previous game to complete before starting the next. Note that since battles run through the harness should not be blocking to begin with (battles naturally wait for players to make their decisions, but the AI's should be making decisions pretty much immediately), this mode is not expected to have large performance benefits over the default sequential mode and may require additional memory (run with node --max-old-space-size=<SIZE> --stack-size=<SIZE> if you encounter issues).

TODO: Add support for running battles in --parallel on muliple cores with worker_threads.